Temporally encoding a static spatial image
System and techniques for temporally encoded static spatial images are described herein. A static spatial image may be obtained. Here, the static spatial image defines pixel values over an area. A scan path may be selected. Here, the scan path defines a path across the area of the static spatial image. A window is scanned (e.g., moved or slid) along the scan path on the static spatial image to produce changes in a portion of the window over time. The changes in the portion of the window are recorded along with respective times of the changes.
1. A system for temporally encoding static spatial images, the system comprising:
electronic circuitry; and
a memory including instructions that, when executed by the electronic circuitry, cause the electronic circuitry to:
obtain a static spatial image, the static spatial image defining pixel values over an area;
select a scan path, the scan path defining:
a path across the area of the static spatial image; and
a duration path, the duration path being a non-linear function that defines progression of the scan path in time;
scan a window in accordance with the scan path on the static spatial image to produce changes in a portion of the window over time; and
record the changes in the portion of the window with respective times of the changes.
2. The system of claim 1 , wherein the window is a single pixel and the portion of the window is the single pixel.
3. The system of claim 1 , wherein to record the changes in the portion of the window with respective times of the changes includes the electronic circuitry to produce a spike at the respective times.
4. The system of claim 3 , wherein the instructions cause the electronic circuitry to:
communicate the spikes to a spiking neural network (SNN) for pattern detection; and
return a pattern detection output from the SNN in response to communicating the spikes.
5. The system of claim 4 , wherein the pattern detection is a multi-part pattern detection.
6. The system of claim 5 , wherein the instructions cause the electronic circuitry to:
receive the spikes as a spike set at an output neuron of the SNN, the spike set meeting a threshold to produce an output spike at the output neuron;
produce an inhibitory spike at the output neuron in response to the spike set meeting the threshold, the inhibitory spike directed to a set of other output neurons;
receive the inhibitory spike at a member of the set of other output neurons; and
inhibit spiking of the member in response to receipt of the inhibitory spike.
7. The system of claim 3 , wherein the spike encodes the changes as a difference from a last state of the portion of the window.
8. The system of claim 1 , wherein each pixel of the static spatial image along the scan path is randomly assigned a fixed time to register a spike within a time period.
9. A method for temporally encoding static spatial images, the method comprising:
obtaining a static spatial image, the static spatial image defining pixel values over an area;
selecting a scan path, the scan path defining:
a path across the area of the static spatial image; and
a duration path, the duration path being a non-linear function that defines progression of the scan path in time;
scanning a window in accordance with the scan path on the static spatial image to produce changes in a portion of the window over time; and
recording the changes in the portion of the window with respective times of the changes.
10. The method of claim 9 , wherein the window is a single pixel and the portion of the window is the single pixel.
11. The method of claim 9 , wherein recording the changes in the portion of the window with respective times of the changes includes producing a spike at the respective times.
12. The method of claim 11 , comprising:
communicating the spikes to a spiking neural network (SNN) for pattern detection; and
returning a pattern detection output from the SNN in response to communicating the spikes.
13. The method of claim 12 , wherein the pattern detection is a multi-part pattern detection.
14. The method of claim 13 , comprising:
receiving the spikes as a spike set at an output neuron of the SNN, the spike set meeting a threshold to produce an output spike at the output neuron;
producing an inhibitory spike at the output neuron in response to the spike set meeting the threshold, the inhibitory spike directed to a set of other output neurons;
receiving the inhibitory spike at a member of the set of other output neurons; and
inhibiting spiking of the member in response to receipt of the inhibitory spike.
15. The method of claim 11 , wherein the spike encodes the changes as a difference from a last state of the portion of the window.
16. The method of claim 9 , wherein each pixel of the static spatial image along the scan path is randomly assigned a fixed time to register a spike within a time period.
17. At least one non-transitory machine readable medium including instructions for temporally encoding static spatial images, the instructions, when executed by a machine, cause the machine to perform operations comprising:
obtaining a static spatial image, the static spatial image defining pixel values over an area;
selecting a scan path, the scan path defining:
a path across the area of the static spatial image; and
a duration path, the duration path being a non-linear function that defines progression of the scan path in time;
scanning a window in accordance with the scan path on the static spatial image to produce changes in a portion of the window over time; and
recording the changes in the portion of the window with respective times of the changes.
18. The at least one non-transitory machine readable medium of claim 17 , wherein the window is a single pixel and the portion of the window is the single pixel.
19. The at least one non-transitory machine readable medium of claim 17 , wherein recording the changes in the portion of the window with respective times of the changes includes producing a spike at the respective times.
20. The at least one non-transitory machine readable medium of claim 19 , wherein the operations comprise:
communicating the spikes to a spiking neural network (SNN) for pattern detection; and
returning a pattern detection output from the SNN in response to communicating the spikes.
21. The at least one non-transitory machine readable medium of claim 20 , wherein the pattern detection is a multi-part pattern detection.
22. The at least one non-transitory machine readable medium of claim 21 , wherein the operations comprise:
receiving the spikes as a spike set at an output neuron of the SNN, the spike set meeting a threshold to produce an output spike at the output neuron;
producing an inhibitory spike at the output neuron in response to the spike set meeting the threshold, the inhibitory spike directed to a set of other output neurons;
receiving the inhibitory spike at a member of the set of other output neurons; and
inhibiting spiking of the member in response to receipt of the inhibitory spike.
23. The at least one non-transitory machine readable medium of claim 19 , wherein the spike encodes the changes as a difference from a last state of the portion of the window.
24. The at least one non-transitory machine readable medium of claim 17 , wherein each pixel of the static spatial image along the scan path is randomly assigned a fixed time to register a spike within a time period.